Capability · Convert

Import anything — not just spreadsheets.

Images and PDFs via OCR, audio via transcription, HTML and URLs via ETL. The extracted records are shaped straight into your target schema — the engine behind “paste anything, drop a doc”.

Get an API key Read the docs

POST /v1/convert

What you get

Built for regulated workloads

Not every source is a file with headers. Convert extracts structured records from unstructured input — scanned forms and PDFs by OCR, recordings by transcription, web pages and URLs by ETL — using our in-region extractor, then shapes the result into your schema. Because unstructured input is full-data by nature, convert runs on the PHI-gated full-data tier; a request without the entitlement returns 402. It’s the capability that powers the Workbench’s “paste anything, drop a doc” flow.

Documents

Images & PDFs → OCR

Scanned forms, faxes and PDFs are OCR’d to structured text in-region, then the model shapes the extracted fields onto your schema — no manual re-keying of a referral or an intake form.

Audio

Recordings → transcription

Audio is transcribed in-region and the salient records are lifted into your target fields — useful for turning a dictated note or a consult recording into rows.

Web

HTML & URLs → ETL

Point convert at a URL or paste HTML and it extracts, transforms and structures the content into records — the same shaping step as every other source, just a different front door.

Routing

Media can’t be PHI-routed

our OCR, transcription and ETL upstreams reject X-PHI, so media conversion deliberately cannot carry protected health information — an honest constraint, surfaced rather than hidden.

Try it

A real call, end to end

Point convert at a document or a URL and name the target schema. It extracts, then shapes the records into your fields — the same output shape as a spreadsheet import.

Good to know

Convert is full-data by nature — it needs the PHI entitlement; without it the endpoint returns 402.
Media upstreams (OCR, STT, ETL) reject X-PHI, so converted media can’t be PHI-routed.
Convert needs a configured LLM key; where that key is unset the endpoint is unavailable (503).

curl

curl https://api.adaptivmapr.com/v1/convert \
  -H "Authorization: Bearer $MAPR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "target": "invoices",
    "source": { "type": "pdf", "data": "data:application/pdf;base64,JVBER…" }
  }'

response

{
  "target": "invoices",
  "rows": [
    { "invoice_no": "INV-2043", "issued_on": "2026-05-14",
      "total": "1240.00", "currency": "EUR" }
  ],
  "extractor": "ocr"
}

FAQ

Common questions

Unstructured input has to be read in full to be extracted, so convert is full-data by nature and runs on the PHI-gated path — a request without the entitlement returns 402. It bills the AI tokens it consumes at provider cost × 2 (× 0.5 with your own key), drawn from your prepaid wallet. Schema-only mode is for tabular sources where only headers and a few sample rows need to leave your system.

Not through the media path. our OCR, transcription and ETL upstreams reject X-PHI, so media conversion is general-content only — that limit is enforced upstream, not left to you. For PHI, use the tabular full-data path where the layer-5 call is PHI-routed.

This endpoint. When you paste prose or drop a document into the Workbench, it routes to convert; tabular pastes go to the cascade instead. Same schema-shaping step at the end, different extractor at the front.

Images and PDFs (OCR), audio (transcription), and HTML or a URL (ETL). Each is extracted in-region and then shaped into your target schema by the same model step.

Structural reshape Schema mapping Schema-only vs full-data

Ready when you are

Put any-to-any convert in production — without shipping raw records.

Spin up a key in minutes. Top up a $10 prepaid wallet — every map is a few tokens.

Get an API key See pricing →

$10 minimum to start · pay only for what you map · PHI under BAA coverage